基于三角形周长的暗星全天球自主快速识别

随着星敏感器极限探测星等能力的提高,导航星暗星数目及识别特征库急剧增加,造成星图识别耗时长,误匹配率高。通过介绍传统的三角形星图识别算法原理及不足,结合主星识别算法思想,以主星对星角距和周长为识别特征,提出了一种基于三角形周长的暗星全天球自主快速识别算法。由于9.0Mv全天球导航特征库的构建过于繁琐,为了提高运算速度,采用NVIDIA图像处理器(GPU)并行计算架构,通过CUDA计算比CPU获得了20倍以上的加速。对特征库按周长构造了散列函数,分段存储成若干个子块。识别过程中,对观测三角形周长采用哈希查找法实现子块的快速定位,星角距的识别只在相应的子块内进行,提高了识别速度,增加了识别成功率。用实拍星图分别对基于星角距和周长特征的识别算法进行了验证,实验表明,两种算法均能实现9Mv星的识别,平均识别时间分别为37.7123 s和2.0422 s,后者具有明显的优势,能够大大提高星敏感器的数据更新率。

Fast all-sky autonomous dark star identification algorithm based on triangle perimeter

With the improvement of limiting magnitude detectivity of star sensor,the number of dark stars in guide stars and the recognition feature library are dramatically increased,which lead to more timeconsuming and higher mismatching rate in star-map recognition.By understanding the principle and deficiencies of traditional triangle star map recognition algorithm,and combined with idea of the main star recognition algorithm,a fast all-sky autonomous dark star identification algorithm is proposed,which takes the main star-pair angular distance and triangle perimeter as the recognition features.Since 9.0Mv all-sky guide feature library is too cumbersome to develop,a NVIDIA Graphics Processor Unit (GPU) parallel computing architecture is adopted to improve the processing speed,which shows that the computation by CUDA is more than 20 times faster than that by CPU.A hash function is constructed for the feature library according to the triangle perimeter and is segmented into sub-blocks.In the process of star identification,the hash look-up method is applied for the triangle perimeter to realize the fast positioning of sub-blocks,while the recognition of star angular distance is only implemented within the corresponding sub-block,which improve the recognition speed and the recognition success rate.Finally,based on a series of real star maps,the experiments are made to verify the star recognition algorithm based on star angular distance and the one based on triangle perimeter,which show that two algorithms can both achieve 9.0Mv star recognition,while the average recognition time is 37.7123 s and 2.0422 s,respectively,which show that the algorithm based on triangle perimeter has obvious advantages,and can greatly improve the data update rate of the star sensor.